awygle / ipl3hasher Goto Github PK

How's anybody gonna know how great this is if I don't tell 'em?

From what I can tell, most of round 1007 is static when Y doesn't change. As data_next, X only impacts 5 words of the state.

So we could implement a truncated round 1007 that only calculates state[10:11] and state[13-15]. Round 1008 is already terminates early and only calculates state[0:9]. Interestingly, state[12] is actually fixed.

Currently every time you re-start the program it starts from y==0. This is obviously dumb, we should support resuming from a different Y value so we can partition the search space over multiple times/people.

This was sheer laziness when I was writing the thing, there's no reason this should be the case.

Several CICs have the same seed but different hashes. We can look for all of these hashes at once and terminate if we find any of them. This should be the default but we should still be able to target just one if we want.

I already wrote this, I just need to pull it into this repo and make it a second binary.

Someone on the N64Brew discord threatened to learn from this code - I should make it less of a Superfund site.

I know what [ 0x0, 0x0 ] means, but nobody else does, or should. Make this clearer.

Currently big-endian is assumed, but all the mechanics are in place to handle little endian except hooking it up to the CLI options.

We can't really support middle-ended with byteorder, but oh well.

In finalise, checksum_hi and checksum_lo are actually completely independent. It's wasteful to calculate both halves before checking.

Whatever half is cheaper should be calculated first and checked against target, only continuing when at least one lane matches.

Hard to tell which is cheaper without testing. checksum_hi only requires 17 multiplies but each multiplies depends on both the previous multiply and the rotate + add. checksum_lo's two chains might have 16 multiplies each but they are actually independent of each other.

CPU hasher could use AVX2 to do 8 hashes at once.